Skip to content
No description, website, or topics provided.
Branch: master
Clone or download
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Type Name Latest commit message Commit time
Failed to load latest commit information.

Build Status codecov


BERT is a powerful workflow for the automated deimmunization of proteins. BERT is based on three modular tools: EvoCS, MHCBoost or alternatively NetMHCIIPan and CoReM. Please refer to our wiki for details: Software.

Quick Setup for users

  1. $ git clone
  2. $ python install
  3. Installation of NetMHCIIpan (
  4. $ BERT

We are unfortunately not allowed to ship NetMHCIIPan, but for academic users it's free of charge. If you have any issues with the setup, please don't hesitate to contact us on .


The CLI - Command Line Interface

$ deimmunization --help
  -in, --input_msa <arg>          file path to msa.txt  [required]
  -rs, --reference_sequence <arg> titlestring of reference sequence
  -t, --majority_threshold <val>  value between 0 and 1 to decide the
                                  consensus value  [required]
  -a, --mhc_alleles <arg>         Input the MHC allele in the following form:
                                  MHC-I: e.g. XXX-000 
                                  MHC-II: e.g. DRB1_0101
                                  It is also possible (and recommended) to
                                  choose multiple alleles. The alleles have to
                                  be comma separated. 
                                  e.g.; DRB1_0101,
                                  DRB1_0202, DRB1_0303, ...  [required]
  -mp, --mhc_ii_pan <arg>         path to netMHCIIpan  [required]
  -mt, --number_mutations <val>   the number of mutation which should be
                                  introduced in the sequence  [required]
  -pos, --pos_to_check <val>      the number of positions to check to
                                  introduce a mutation in each iteration
  -pdb, --pdb_file <arg>          path to pdb file  [required]
  -cha, --pdb_chain <arg>         Chain in the pdb file  [required]
  -w, --weight <val>              Weighting factor between deimmunization and
                                  ddG. A low value favors low immunity and a
                                  high valuefavors a stable protein
  -frq, --allele_frequency <val>  Frequency of the alleles in the target
                                  population. The frequencies have to be comma
                                  separated  [required]
  --help                          Show this message and exit.


Example command:

BERT -in <MSApath> -rs <ref.seqMSA> -t 0.5 -a DRB1_0301,DRB1_0701,DRB1_1501 -mp <path_netMHCIIPan> -pdb <path.pdb> -pos 2 -mt 2 -cha B -w 10 -frq 0.099,0.109,0.0967    

You can use the following MSA, PDB and sequence files shipped with this program: Example MSA: 'data/Isomerase_90_similarity.clustal_num'
Example PDB: 'data/2wcv.pdb'
Example Reference Sequence: '2WCV_Paris'

Performance and Results

BERT has been applied to team iGEM Paris-Bettencourt's FucU isomerase protein. The results are documented on our wiki: Collaboration. Moreover, BERT was used to deimmunize BoNTC. Details can be found on our software page: Software.


Our tool is made available under the MIT License.


Team iGEM 2018 Tübingen
Steffen Lemke
Lukas Heumos
Alexander Röhl

You can’t perform that action at this time.